Do MysQL Query will less re-read about the same table.

marrowyung
marrowyung used Ask the Experts™
on
Dear all,

I read this: http://blog.sqlauthority.com/2014/02/25/mysql-profiler-a-simple-and-convenient-tool-for-profiling-sql-queries/

and I am not sure why this one:

SELECT DISTINCT p.maker_id
FROM products AS p
JOIN models m ON p.model_id = m.model_id
WHERE m.model_type = 'PC'
AND NOT EXISTS (
SELECT p2.maker_id
FROM products AS p2
JOIN models m2 ON p2.model_id = m2.model_id
WHERE m2.model_type = 'Laptop'
AND p2.maker_id = p.maker_id
);

Open in new window


can be converted to :

SELECT p.maker_id
FROM products AS p
JOIN models AS m ON p.model_id = m.model_id
WHERE m.model_type IN ('PC', 'Laptop')
GROUP BY p.maker_id
HAVING COUNT(CASE WHEN m.model_type = 'PC' THEN 1 END) > 0
AND COUNT(CASE WHEN m.model_type = 'Laptop' THEN 1 END) = 0;

Open in new window



with the same result?

what is the

HAVING COUNT(CASE WHEN m.model_type = 'PC' THEN 1 END) > 0
AND COUNT(CASE WHEN m.model_type = 'Laptop' THEN 1 END) = 0;

Open in new window


is about?
Comment
Watch Question

Do more with

Expert Office
EXPERT OFFICE® is a registered trademark of EXPERTS EXCHANGE®
SELECT DISTINCT p.maker_id
FROM products AS p
JOIN models m ON p.model_id = m.model_id
WHERE m.model_type = 'PC'
AND NOT EXISTS (
SELECT p2.maker_id
FROM products AS p2
JOIN models m2 ON p2.model_id = m2.model_id
WHERE m2.model_type = 'Laptop'
AND p2.maker_id = p.maker_id
);

Meaning of this is - required distinct maker_ id where model type is only PC not Laptop

in new query

GROUP BY p.maker_id
HAVING COUNT(CASE WHEN m.model_type = 'PC' THEN 1 END) > 0
AND COUNT(CASE WHEN m.model_type = 'Laptop' THEN 1 END) = 0;

GROUP BY - use for distinct data

HAVING COUNT(CASE WHEN m.model_type = 'PC' THEN 1 END) > 0  - use for make sure model type is PC


AND COUNT(CASE WHEN m.model_type = 'Laptop' THEN 1 END) = 0; - use to make sure that it will not contain laptop as model type
Most Valuable Expert 2017
Distinguished Expert 2018
Commented:
The first query basically says find all records where the products contains 'PC' and where there isn't another record for the same manufacturer that makes a laptop.

In otherwords only find manufacutures of PC's

The second query obviously does the same thing but uses a slightly different technique.

This query groups all records that have the same manufacturer id together where the manufacture supplies either PC's or laptops and then sums (COUNT) the number of each device a particular manufacture makes.

So what could come out of the query - before the having is applied is

+----------+----------------------------+--------------------------------+
| MakerId  | Count(m.modelType  = 'PC') |  COUNT(m.modelType = 'Laptop') |
+----------+----------------------------+--------------------------------+
|  1       |           1                |            1                   |
|  2       |           1                |            0                   |
|  3       |           0                |            1                   |
+----------+----------------------------+--------------------------------+

Open in new window

The HAVING clause then exclude the records where PC is  0 or LAPTOP is > 0

Or to put it another way

Includes where PC is > 0
HAVING COUNT(CASE WHEN m.model_type = 'PC' THEN 1 END) > 0

Open in new window

AND
where Laptop = 0
COUNT(CASE WHEN m.model_type = 'Laptop' THEN 1 END) = 0

Open in new window

Leaving you with
+----------+----------------------------+--------------------------------+
| MakerId  | Count(m.modelType  = 'PC') |  COUNT(m.modelType = 'Laptop') |
+----------+----------------------------+--------------------------------+
|  2       |           1                |            0                   |
+----------+----------------------------+--------------------------------+

Open in new window

marrowyungSenior Technical architecture (Data)

Author

Commented:
basically is using the HAVING COUNT or count in this way can minimize the re read problem but is it just because of the having or becauase of the simple calculate in the where cause instead of "NOT EXISTS " ?

this will make the MySQL faster ?

Which one is the key part ?
Exploring ASP.NET Core: Fundamentals

Learn to build web apps and services, IoT apps, and mobile backends by covering the fundamentals of ASP.NET Core and  exploring the core foundations for app libraries.

Most Valuable Expert 2017
Distinguished Expert 2018

Commented:
It is a combination - grouping of records on like makerid and then eliminating the records you don't want based on the count values.
marrowyungSenior Technical architecture (Data)

Author

Commented:
but grouping of records on like makerid don't speed up thing, right ?


so eliminating the records you don't want based on the count values is the key part ?
grouping of records on like makerid will speed up thing .
marrowyungSenior Technical architecture (Data)

Author

Commented:
then the inner join (to  replace the sub query) is not that efficent enought than that?

Usually the key thing should be that any thing in the where cause should be index to avoid table scan, this should provide much performance gain than this kind of group by, right?

Do more with

Expert Office
Submit tech questions to Ask the Experts™ at any time to receive solutions, advice, and new ideas from leading industry professionals.

Start 7-Day Free Trial